MBP 15'' late 2016 & liquid metal thermal paste tests

[ionantonak]

Hello everyone, this is my first post in these forums so if I make any mistakes please let me know.

As the title suggests I had the idea to swap the stock thermal compound of my Mac with the thermal grizzly conductonaut. The model that I have has the Radeon pro 460 and the 6820HQ.

To be honest I didn't have any temperature problems but as an engineer I always like to experiment with stuff and teardown the devices that I own. I do realise that Liquid metal isn't happy when it hangs out with aluminium and that it can kill board connections but yolo and since no-one else had done it for the 2016 I decided to go with it.

Disassembly (factory thermal paste):


Cleaning it up:


Applying the 33+ electrical tape to avoid any shorts and the liquid metal itself:


(Note that this bead of liquid metal was used for the CPU, GPU and the heatsinks)

So I spread the thermal compound and plugged everything back together; btw apple keeps on making the laptops more difficult to take apart.

And the results are in... Not that exciting to be honest 4-5 degrees Celsius at max. At this point I should point out that I didn't have a good base line, I used MKBHD's 8k video while having the automatic graphics switching off (using the AMD gpu) and with the fans set at either default or at 100% using istat menus.

Stock:
default fan settings - 82 degrees Celsius
100% fan speed - 60 degrees Celsius

Conductonaut:
default fan settings - 74 degrees Celsius
100% fan speed - 58 degrees Celsius

Now I moved to Windows 10 (I use bootcamp) and used Furmark and prime95 to stress test everything with the fans always at 100% while using the Conductonaut.

Furmark:
CPU - peak package temp 79 degrees Celsius
GPU - peak package temp 62 degrees Celsius

Prime95:
CPU - peak package temp 91 degrees Celsius
GPU - peak package temp 66 degrees Celsius

Furmark & Prime95 (small FFTs):
CPU - peak package temp 79 degrees Celsius
GPU - peak package temp 73 degrees Celsius

I will try to update the post with more info as soon as I have time and to be honest I will probably change the paste back to something like this as it is not worth the risk for these temps.

If you have Windows 10 and time please run the same benchmarks for 5 minutes and post the results to see if it actually makes a bigger difference.

(Note that throughout my testing the room temp didn't change that much as I was monitoring the palm rest sensor 2 at all the time and it was at a ~20 degrees Celsius. I know that this is not the proper way to do it but who has time for science )

 

[jerryk]

Wow, you put a lot of paste on those surfaces.

I just use a tiny bit of Artic Silver or other compound. Just enough to fill in the irregularity of the surfaces.

 

[ionantonak]

Wow, you put a lot of paste on those surfaces.

I just use a tiny bit of Artic Silver or other compound. Just enough to fill in the irregularity of the surfaces.

That is the thermal paste coming from the factory believe it or not, I swapped it for the liquid metal

 

[jerryk]

That is the thermal paste coming from the factory believe it or not, I swapped it for the liquid metal

I was referring to the liquid metal.

 

[ionantonak]

I was referring to the liquid metal.

That amount was spread over the CPU and GPU and the heatsink for both as thinly as possible plus I used the included q-tip which absorbed some of it as well, I should have taken more pictures to be honest but still it wouldn't explain the temps

 

[bartvk]

Very, very nice work. Especially grateful for the detailed photos.

Also, I think the result is pretty good. Basically you got a ten percent improvement, that's very cool. Makes me wonder why Apple didn't do so.

Edit: ah, I get it... While applying the liquid metal, you were using the tape to prevent spreading outside the lid. I can see why Apple doesn't use liquid metal, I bet this causes yield issues.

 

[jerryk]

I wonder what temps this system would get with a good quality thermal paste properly applied. I remember I got the temps down 10C+ on a Dell i7 laptop by simply removing the factory applied thermal paste, cleaning it up (it was literally dripping down the sides of the main board), and using a dab of Arctic Silver paste.

 

[ionantonak]

Very, very nice work. Especially grateful for the detailed photos.

Also, I think the result is pretty good. Basically you got a ten percent improvement, that's very cool. Makes me wonder why Apple didn't do so.

Edit: ah, I get it... While applying the liquid metal, you were using the tape to prevent spreading outside the lid. I can see why Apple doesn't use liquid metal, I bet this causes yield issues.

Thanks very much, yeah the whole liquid metal idea is quite dodgy on it own. It makes sense that no company is using it at the moment, as a tiny spill can take out a complete board or CPU.

I wonder what temps this system would get with a good quality thermal paste properly applied. I remember I got the temps down 10C+ on a Dell i7 laptop by simply removing the factory applied thermal paste, cleaning it up (it was literally dripping down the sides of the main board), and using a dab of Arctic Silver paste.

I did the same thing on an Alienware 15 two years ago and noticed a huge difference. I believe that in my case and in general MacBooks are not limited that much by their thermal compound but by the cooler itself. There is no way that ~80W can be dissipated from that tiny heat pipe and the small fins, the surface area is just not enough. On the other hand it makes some sense to cheap out on the thermal paste cause when you ship millions of them even a 0.10 £ or $ will lead to loads of more profit

 

[Calby]

Not to bee that guy but... do you have the logicboard on a carpet?
That's really ESD dangurus.

 

[Mindinversion]

On the bright side, factory paste application looks really good. I've seen some teardown where it looked like someone used an entire tube of thermal paste on a single laptop

 

[ionantonak]

Not to bee that guy but... do you have the logicboard on a carpet?
That's really ESD dangurus.

no worries, it is on a microfibre cloth. Not the best option but I did ground myself before opening it up which helps

 

[jerryk]

I did the same thing on an Alienware 15 two years ago and noticed a huge difference. I believe that in my case and in general MacBooks are not limited that much by their thermal compound but by the cooler itself. There is no way that ~80W can be dissipated from that tiny heat pipe and the small fins, the surface area is just not enough. On the other hand it makes some sense to cheap out on the thermal paste cause when you ship millions of them even a 0.10 £ or $ will lead to loads of more profit

But, hey look at the skinny design. Slimness at all costs!

 

[Sanpete]

I believe that in my case and in general MacBooks are not limited that much by their thermal compound but by the cooler itself. There is no way that ~80W can be dissipated from that tiny heat pipe and the small fins, the surface area is just not enough. On the other hand it makes some sense to cheap out on the thermal paste cause when you ship millions of them even a 0.10 £ or $ will lead to loads of more profit

The new MBPs haven't shown any issues with heat, so that may be why there's no additional paste.

 

[ionantonak]

The new MBPs haven't shown any issues with heat, so that may be why there's no additional paste.

I never said that it overheated or anything, I just tried something that I could not find online. In any case with thermal compounds it is not about quantity, it is about quality, application and thermal conductivity. In most cases something like Arctic Silver 5 or the Gelid extreme is much better than the factory stuff

 

[Sanpete]

I never said that it overheated or anything, I just tried something that I could not find online. In any case with thermal compounds it is not about quantity, it is about quality, application and thermal conductivity. In most cases something like Arctic Silver 5 or the Gelid extreme is much better than the factory stuff

I understand. My point is that there's no basis for supposing that the new MBPs are excessively thin or that they skimp on thermal paste based on their actual heat management, which is excellent.

 

[SesameWasher]

AFAIK, boot camp usually takes more power due to crappy drivers. I wonder if the temperature will improves if you can run your stress tests in macOS.

Well there is Prime95 available for Macs, as well as Linpack on the Intel website. Although I can't find (nor really care) anything that stress the GPU.

https://www.mersenne.org/download/
https://software.intel.com/en-us/mk...e-intel-optimized-linpack-benchmark-for-macos

I swaped the thermal paste on my old MacBook years ago, it never goes above 76 Celsius when I tried Linpack, and mostly 40-50 Celsius under idle. A huge jump compared to 60-70 Celsius idle and even 90-110 Celsius stress.
That being said, it's a dual-core processor, and the thermal paste back then were pretty much dried off.

If you still interested to test the CPU out, I'll be glad to see it.

 

[Queen6]

Those interested in this thread will also likely be interested in this one on MR:

Automated-tool-to-reveal-throttling-and-overheating-github


Q-6

 

[SesameWasher]

Those interested in this thread will also be interested in this one on MR:

Automated-tool-to-reveal-throttling-and-overheating-github


Q-6

which, I don't see anyone participated for current generations...

 

[Queen6]

which, I don't see anyone participated for current generations...

Then someone needs to get the ball rolling

Q-6

 

[SesameWasher]

Then someone needs to get the ball rolling

Q-6

I sure hope so, too bad I don't have one. The old one that I have works thermally amazing.

 

[Queen6]

I sure hope so, too bad I don't have one. The old one that I have works thermally amazing.

2016 Model onwards is the best MBP regarding thermal design, so I expect that explains the lack of participants. Previous generations certainly did have issues related to throttling, equally model to model and spec dependent.

16/17 MBP doesn't work for me, so on a different platform these days, likely try the 2018, equally I don't have high hopes...

Q-6

 

[SesameWasher]

2016 Model onwards is the best MBP regarding thermal design, so I expect that explains the lack of participants. Previous generations certainly did have issues related to throttling, equally model to model and spec dependent.

16/17 MBP doesn't work for me, so on a different platform these days, likely try the 2018, equally I don't have high hopes...

Q-6

I can't just guess myself. On the other hand, I do saw reviews from notebookcheck, but unfortunately, they only do stress test under windows, which boot camp drivers were famously known to suck up more power than it needs. 91 Celsius seems reasonable under boot camp, though.

For the Haswell series (4th-Gen), that's mostly because the voltage regulator. It was promised to have a "smarter power regulation", but instead it failed and makes chips whole lot warmer.

About the 2018 thingy, I'm sure that has more to do with Intel than Apple. In fact, many PC builders aren't happy either, because it took so long to finally got hex-core processors rolled out on the mainstream platforms. That being said, it could be another challenge for the thermal system once the Coffee Lake-H shows up.

 

[Queen6]

I can't just guess myself. On the other hand, I do saw reviews from notebookcheck, but unfortunately, they only do stress test under windows, which boot camp drivers were famously known to suck up more power than it needs. 91 Celsius seems reasonable under boot camp, though.

For the Haswell series (4th-Gen), that's mostly because the voltage regulator. It was promised to have a "smarter power regulation", but instead it failed and makes chips whole lot warmer.

About the 2018 thingy, I'm sure that has more to do with Intel than Apple. In fact, many PC builders aren't happy either, because it took so long to finally got hex-core processors rolled out on the mainstream platforms. That being said, it could be another challenge for the thermal system once the Coffee Lake-H shows up.

Apple will always tend to lag as they place form over function to please the masses. Certainly interested in Hex Core, equally only in a notebook that can handle the performance with a strong GPU and uncompromising usability, which currently rules Apple out...

Q-6

 

[SesameWasher]

Apple will always tend to lag as they place form over function to please the masses. Certainly interested in Hex Core, equally only in a notebook that can handle the performance with a strong GPU and uncompromising usability, which currently rules Apple out...

Q-6

Hex-core mobile processors are something set in stone for Intel, it will sure have that huge performance jump like 8700K vs. 7700K.

Another thing that people tend to forgot is the GPU inside a MacBook Pro never went higher than mid-tier ever since Apple switched to Intel. The major difference is perhaps how people complained today compared how they loved the classic ones. That being said, everything that's a "high-end" GPU would required a design that's even thicker than those unibody MacBook Pros, if you want to achieve the full performance out of it.

 

[Queen6]

Hex-core mobile processors are something set in stone for Intel, it will sure have that huge performance jump like 8700K vs. 7700K.

Another thing that people tend to forgot is the GPU inside a MacBook Pro never went higher than mid-tier ever since Apple switched to Intel. The major difference is perhaps how people complained today compared how they loved the classic ones. That being said, everything that's a "high-end" GPU would required a design that's even thicker than those unibody MacBook Pros, if you want to achieve the full performance out of it.

No doubt on the 8th Gen. GPU'S Apple's pretty much out the game for those with higher needs. Many do appreciate the "balance" of portability & performance, equally for many the new MBP is overly compromised. Personally I would far prefer a more usable, capable notebook as being needlessly thinner has only resulted in ceasing to use the Mac professionally not needing or wanting the compromises the new design brings.

Q-6